JPH04183561A - Expert system for decision of process state - Google Patents

Abstract

PURPOSE:To execute the management of process effectively, by executing the function deciding a process state automatically by a computer, with 1st, 2nd inference mechanisms with the use of a management chart. CONSTITUTION:In the management chart prepared from various kinds of data measured in the process which becomes a management object, a 1st inference mechanism 6 which decides the trend of the variation of the management chart is first provided. Then, the state of process is inferred by a 2nd inference mechanism 8 from the inference result with respect to the trend obtained by this 1st inference mechanism 6.

Description

DETAILED DESCRIPTION OF THE INVENTION (Industrial Application Field) The present invention relates to a process status determination expert system that automatically determines process status in a manufacturing process or the like.

(Prior Art) A control chart is one of the effective means of the QC method to measure the dimensions of a workpiece in a certain process, represent the measurement results in a graph, and judge the state of the process. Figure 2 is control chart 1
This is an example of a species-R control chart.

The x-R control chart divides a plurality of measurement data into groups in time order, and represents the average value and variation within each group in a Ma control chart and an R control chart, respectively. The feature of this control chart is that changes in the average value and variation for each group can be seen at a glance, making it easy to understand changes in process status. In managed processes, control charts are always created, and operators judge the process status from the control charts, discover and predict process abnormalities, and take countermeasures.

(Problems with the Prior Art) As described above, control charts are positioned as an important control method of QC methods as an effective means of process control for determining the state of a process and taking countermeasures.

However, while computer software and other products are available for automatically creating control chart graphs from data, skilled operators are required to judge the process status from the created control chart graphs. This is what is currently being done. Therefore, if a skilled person is absent for some reason, the process cannot be adequately managed even if a control chart is created.

An object of the present invention is to solve the above-mentioned problems and provide an expert system for determining process status that can automatically determine the status of a process.

(Means for Solving the Problems) According to the present invention, in a control chart created from various data measured in a process to be managed, a first inference mechanism that first determines the tendency of change in the control chart; A process state determination expert system has been provided that includes a second inference mechanism that infers the state of the process from the inference results regarding trends obtained by the first inference mechanism.

(Example) Hereinafter, specific examples of the present invention will be described. Here, an example will be described in which the state of the process is determined from the processed dimensions of the workpiece.

FIG. 1 shows a processing flow for determining process status. 1 is a workpiece, and 2 is a dimension measuring instrument such as an electric micrometer for measuring the machining dimensions of the workpiece. The output of 2 is sent to an amplification/A-D converter buffer, where it is amplified and further converted into a digital signal in the case of digital signal processing. The average value/variation calculator in step 4 uses the measurement data obtained in step 3 as data for each group, and calculates the average value and variation of the measurement data within the group. The data of the average value (Ma) and the variation (R) obtained in step 4 are stored in the memory 5 as a temporal flow. FIG. 2 is a graph of the data stored in the memory 5, which is usually called an x-R control chart.

Next, the first inference mechanism 6 is activated based on the data stored in the memory 5. The first inference mechanism 6 determines the trend of data, for example, a run (11 in FIG. 2), an upward trend (12 in FIG.
), downward trend (13 in Figure 2), bias (14 in Figure 2)
This is an expert system connected to a first rule base 9 that extracts features such as, for example, if four consecutive pieces of data are on an upward trend, or four or more consecutive pieces of data are on the other hand with respect to the center line. Infer trends in temporal changes in data, such as bias towards (ren). Here, the first rule base 9 used in the first inference mechanism is knowledge that is generalized as a way of viewing control charts, and therefore can be used in common regardless of the purpose of using the control chart. Naturally, different trends such as continuous, upward trend, and bias may occur at the same time, so all detected trend patterns are stored in the memory 7. That is, the memory 7 stores characteristic portions of data trend patterns.

Finally, the second inference mechanism 8 is activated based on the trend pattern data stored in the memory 7. The second inference mechanism is an expert system connected to a second rule base 10 that detects abnormalities in the process from the trend patterns stored in the memory 7. infer the state of the process. For example, if upward trends and downward trends of 4 or more consecutive points repeatedly appear, the process status is normal, but if only upward trends occur many times, an abnormality has occurred. A second rule like this
By inputting the information into the rule base 10 of the first expert system, it becomes possible to determine whether the process status is normal or abnormal based on the trend data obtained by the first expert system. It is also possible to infer the status of processing machines such as grinders. For example, if a downward trend appears frequently and the period becomes short, the grinding wheel may be worn out. Therefore, by inputting this trend into the second rule base IO, it is possible to output a message such as "The grinding wheel is getting worn out. It needs to be replaced" to alert the operator. Become. In this way, the second rule base lO used by the second inference mechanism is knowledge specific to each individual process.

A feature of the present invention is that, by utilizing an expert system, a computer can automatically perform the ability to judge process status from a control chart, which conventionally had to be done by experts. Furthermore, a first inference mechanism connects a first rule base that is common among individual processes, and a second inference mechanism connects a second rule base that depends on each individual process.
The point is that the expert system is composed of the inference mechanism. In other words, by separating rules common to many processes and rules unique to each process, it is possible to realize many process status judgment expert systems with a small amount of development.

(Effects of the Invention) As described above, the present invention basically allows a computer to automatically perform the function of determining the process status using a control chart, thereby making it possible to efficiently manage the process. It became possible. Therefore, in the example, all (two different
Although the description has been made in such a way as to execute a type of expert system, this is done to facilitate the explanation of the inference process, and is not limited to this. Even if the two rule bases 9 and 10 are created as one rule base, the same effect can be obtained by adding a rule called a control rule that determines the order of inference.

In addition, although the x-R control chart was used in the example as a control chart, control charts are created for the purpose of process control as mentioned earlier, and there are also various types of control charts that can be used depending on the purpose. The figure is known. The present invention includes a control chart created based on data collected for the purpose of process management.

[Brief explanation of the drawing]

FIG. 1 is an explanatory diagram of an embodiment showing the configuration of the present invention. 1 is a workpiece, and 2 is a measuring instrument for measuring the machining dimensions of the workpiece. In this example, the external dimensions of the workpiece are used as a proxy characteristic representing the state of the process. 3 is an input buffer that performs preprocessing such as amplifying the data of the measuring instrument and digital conversion, and 4 is 3
A calculator that creates x-R control chart data from the measurement data obtained from 5, 5 is a memory that stores the output of 4, and 6 is 5
7 is a memory that stores the output of 6, and 8 is stored in 7. A second inference mechanism that determines and outputs the state of a process from trend data and a second rule base that determines the state of an IO process is shown. FIG. 2 is an example of a MarR control chart, which is a type of control chart illustrating the contents of the memory 5 shown in FIG. The upper graph shows the average value (Ma), and the lower graph shows the control chart for the variation (R). The horizontal axis is usually the time axis, and time passes from left to right (CLSUCL in the diagram. LCL is a term used for control charts and has the following meaning. CL: Center line UCL: Upper control limit line LCL :Lower control limit line 11-14 are examples of trends in control charts, where 11 is a run, 12 is an upward trend, 13 is a downward trend, and 14 is an example of bias.

Claims (1)

[Claims] Input a control chart created from various data measured within the process to be managed, and use the first input and memorized control chart.
The first inference mechanism that determines the tendency of temporal changes in the control chart and the inference results regarding the trends obtained by the first inference mechanism are inputted with reference to the rule base of A process state judgment expert system comprising a second inference mechanism that infers the state of the process by referring to the rule base of.